Note
Go to the end to download the full example code.
Demonstration of the Factory classes for reliability problems¶
In this example, we show how to use various classes which provide an easy way to create an algorithm to estimate a problem from a ReliabilityBenchmarkProblem. This methods do not set the parameters of the algorithm and do not run it, so that we can set specific settings for a given problem.
import openturns as ot
import otbenchmark as otb
We consider the RP8 problem.
problem = otb.ReliabilityProblem8()
Create a Monte-Carlo algorithm¶
The buildMonteCarlo method creates a ProbabilitySimulationAlgorithm based on MonteCarlo sampling. Before running the algorithm, we set the number of outer iterations based on the setMaximumOuterSampling method. This shows the main utility of the Factory classes.
factory = otb.ProbabilitySimulationAlgorithmFactory()
algo = factory.buildMonteCarlo(problem)
algo.setMaximumOuterSampling(100000)
algo.run()
result = algo.getResult()
result.getProbabilityEstimate()
0.0006900000000000031
Create a FORM algorithm¶
We use the FORM class applied to the problem.
nearestPointAlgorithm = ot.AbdoRackwitz()
algo = otb.FORM(problem, nearestPointAlgorithm)
The FORM object of the otbenchmark module implements
a openturns.FORM
object.
Hence, it has a run method. If specific setting is required,
we can do it now, prior to the call to the run method.
algo.run()
result = algo.getResult()
result.getEventProbability()
0.0006598990293278664
We can compare the previous estimate with the exact probability.
problem.getProbability()
0.0007897927545597477
Create a SORM algorithm¶
The SORM class creates a SORM object.
nearestPointAlgorithm = ot.AbdoRackwitz()
algo = otb.SORM(problem, nearestPointAlgorithm)
algo.run()
result = algo.getResult()
result.getEventProbabilityBreitung()
0.000783711312874782
Create a FORM-IS algorithm¶
The buildFORMIS method of the ProbabilitySimulationAlgorithmFactory class creates a ProbabilitySimulationAlgorithm object, based on the Importance Sampling method using the FORM design point with gaussian importance distribution.
factory = otb.ProbabilitySimulationAlgorithmFactory()
nearestPointAlgorithm = ot.AbdoRackwitz()
algo = factory.buildFORMIS(problem, nearestPointAlgorithm)
algo.run()
result = algo.getResult()
result.getProbabilityEstimate()
0.0005702210872887895
Create a SubsetSampling algorithm¶
algo = otb.SubsetSampling(problem)
algo.run()
result = algo.getResult()
result.getProbabilityEstimate()
0.0007323000000000017
Create a LHS algorithm¶
algo = otb.LHS(problem)
algo.run()
result = algo.getResult()
result.getProbabilityEstimate()
0.0
Total running time of the script: (0 minutes 6.196 seconds)